The promise and perils of synthetic data | TechCrunch

Synthetic data is increasingly hard to come by, as new, real data is hard to obtain.

The market for annotated AI training services is estimated to be worth $10.34 billion in the next 10 years .

But why does AI need data in the first place — and what kind of data does it need? And can this data be replaced by synthetic data?.

Gartner predicts 60% of the data used for AI and analytics projects this year will be synthetically generated.

Synthetic data can be used to generate training data in a format that’s not easily obtained through scraping (or even content licensing) It suffers from the same “garbage in, garbage out” problem as all AI .

Complex models such as OpenAI 's o1 could produce harder-to-spot hallucinations in synthetic data.

OpenAI CEO Sam Altman once argued that AI will someday produce synthetic data good enough to effectively train itself.

No major AI lab has released a model trained on synthetic data alone.

At least for the foreseeable future, it seems we’ll need humans in the loop somewhere to make sure a model’s training doesn’t go awry.